pandit ☸️ pandas utils

Pandas with some cool additional features

Installation and usage

pip install pandit

import pandas as pd, import pandit
# or 
import pandit as pd

df=pd.read_tsv(path)
df.sieve(x=3).show()
#Pandas behaves normally otherwise

If credentials are needed:

import credentials # you manage that part
assert credentials.gsheet # credential dict in https://docs.gspread.org/en/latest/oauth2.html
assert credentials.dropbox 
pd.credentials = credentials

sieve

df.sieve(column1=value1, columns2=value2)
# returns df rows where column equals value - if value is not a list, otherwise:
df.sieve(column3=[value1,value2])
# returns df rows where column is value1 or value2; use [[value1,value2]] to match lists
# It's like pd.query but with a pythonic syntax instead of the sql string.

show

df.show() # shows multiple rows column by column (one line per column) with nice formatting, one line per column
# ideal for inspecting NLP datasets
df.rshow(n) # random sample of size n (default is 20)

Also:

df.bold_max()

bold max float values df.bold_max().to_latex()

pd.read_tsv

read_csv with sep='\t' for lazy persons

pd.read_jsonl

pd.read

df.read_{extension} where extension is extracted from the input path (.csv = read_csv)

pd.read_wandb(project_name)

df.drop_constant_column

drop columns that are constant

df.to_dropbox(path, format=None, token=None,**kwargs)

Save dataframe to dropbox

df.to_sheets(id,sheet_name,credential=None, include_index=False)

Save dataframe to sheets

df.undersample(column='label',sampling_strategy='auto',random_state=None,replacement=False)